Table of contents
Optimising your AI chatbot
Congratulations! Your AI chatbot is now live and ready to help your customers! But the work does not end here. Now that your chatbot is talking to your customers, it’s time to optimise your chatbot.
Optimising your chatbot is a never-ending process in which you are constantly analysing conversations and adding new knowledge to your chatbot. The best way to optimise the Artificial Intelligence of your chatbot is to take a good look at the Mismatches.
When a customer asks your chatbot a question to which it doesn’t yet know the answer, your chatbot will create a Mismatch. Because the chatbot couldn’t match the question to an answer, it will forward the question to an available agent, so your customer will still receive the answer they’re looking for. The unknown question will be saved though, so you can teach your chatbot this new knowledge.
In your Watermelon platform you can find an overview of all the Mismatches your chatbot has encountered. From here you can analyse the unknown questions and determine if they are relevant to your chatbot. If they’re not, you can simply delete them and clean up your Mismatches.
If a Mismatch is relevant to your chatbot, you can add this new question to your chatbot by either adding it as a new variation to an existing dataset or by creating a completely new dataset and thinking of new variations for this question.
By regularly adding Mismatches to the knowledge of your chatbot, you can make sure your chatbot is continuously developing and can help your customer better and better everyday!
Step 5. Best practices for a successful AI chatbot
We would like to share a few tips and tricks to help you make the most out of your artificially intelligent chatbot.